How are LLMs able to solve millennium prize problems?
retained text
I dont know a lot about AI and am not sure if my understanding is outdated or inaccurate, so please feel free to correct me or clarify anything I'm misunderstanding. How are these models able to solve such problems? I guess a more general question is how these models are able to generate a novel solution to any problem at all? I am under the assumption that they can only exhibit patterns and behavior that are present in the massive amount of data they are trained on, so when it comes to these problems that supposedly no one has solved before, and therefore there couldn't be any data to train on in regards to a solution, then what is allowing them to do so? Edit: I can't respond to everyone but I am reading every post and I appreciate everyone taking the time to inform me.
document fetched 2026-10-14 00:28 · canonical https://reddit.com/r/OpenAI/comments/1x4qpcq/how_are_llms_able_to_solve_millennium_prize